Gaucher Disease: The Origins of the Ashkenazi Jewish N370S and 84GG Acid β-Glucosidase Mutations

Am. J. Hum. Genet. 66:1821–1832, 2000

1821

Gaucher Disease: The Origins of the Ashkenazi Jewish N370S and 84GGAcid b-Glucosidase MutationsGeorge A. Diaz,1,2,* Bruce D. Gelb,1,2,* Neil Risch,4 Torbjoern G. Nygaard,5 Amos Frisch,6,8

Ian J. Cohen,7,8 Clara Sa Miranda,9 Olga Amaral,9 Irene Maire,10 Livia Poenaru,11

Catherine Caillaud,10 Moishe Weizberg,1 Pram Mistry,1,3 and Robert J. Desnick1,2

Departments of 1Human Genetics, 2Pediatrics, and 3Medicine, Mount Sinai School of Medicine, New York; 4Department of Genetics, StanfordUniversity, Stanford; 5Department of Neuroscience, UMDNJ-New Jersey Medical School, Newark; 6Felsenstein Medical Research Center,Rabin Medical Center, and 7Schneider Children’s Medical Center of Israel, Petah Tikva, Israel; 8Sackler Faculty of Medicine, Tel AvivUniversity, Tel Aviv; 9Instituto de Genetica Medica, Porto, Portugal; 10Hopital Debrousse, Lyon; and 11Laboratoire de Genetique, HopitalCochin, Paris

Type 1 Gaucher disease (GD), a non-neuronopathic lysosomal storage disorder, results from the deficient activityof acid b-glucosidase (GBA). Type 1 disease is panethnic but is more prevalent in individuals of Ashkenazi Jewish(AJ) descent. Of the causative GBA mutations, N370S is particularly frequent in the AJ population, ( ),q ∼ .03whereas the 84GG insertion ( ) occurs exclusively in the Ashkenazim. To investigate the genetic history ofq ∼ .003these mutations in the AJ population, short tandem repeat (STR) markers were used to map a 9.3-cM regioncontaining the GBA locus and to genotype 261 AJ N370S chromosomes, 60 European non-Jewish N370S chro-mosomes, and 62 AJ 84GG chromosomes. A highly conserved haplotype at four markers flanking GBA (PKLR,D1S1595, D1S2721, and D1S2777) was observed on both the AJ chromosomes and the non-Jewish N370S chro-mosomes, suggesting the occurrence of a founder common to both populations. Of note, the presence of differentdivergent haplotypes suggested the occurrence of de novo, recurrent N370S mutations. In contrast, a differentconserved haplotype at these markers was identified on the 84GG chromosomes, which was unique to the AJpopulation. On the basis of the linkage disequilibrium (LD) d values, the non-Jewish European N370S chromosomeshad greater haplotype diversity and less LD at the markers flanking the conserved haplotype than did the AJ N370Schromosomes. This finding is consistent with the presence of the N370S mutation in the non-Jewish Europeanpopulation prior to the founding of the AJ population. Coalescence analyses for the N370S and 84GG mutationsestimated similar coalescence times, of 48 and 55.5 generations ago, respectively. The results of these studies areconsistent with a significant bottleneck occurring in the AJ population during the first millennium, when thepopulation became established in Europe.

Introduction

The unique demographic history of the Ashkenazi Jew-ish (AJ) population and the occurrence of several lyso-somal storage disorders at high frequency in this grouphave attracted considerable interest and debate concern-ing the mechanism(s) underlying the prevalence of thesediseases among the Ashkenazim (Chase and McCusick1972; Myrianthopolous et al. 1972; Goodman 1979;Zlotogora et al. 1988; Bonne-Tamir and Adam 1992;Motulsky 1995; Zoosmann-Diskin 1995). The recentavailability of high-density genetic maps has facilitated

Received November 17, 1999; accepted for publication March 14,2000; electronically published April 21, 2000.

Address for correspondence and reprints: Dr. Robert J. Desnick,Department of Human Genetics, Mount Sinai School of Medicine,New York, NY 10029. E-mail: [email protected]

* These authors contributed equally to this work.q 2000 by The American Society of Human Genetics. All rights reserved.

0002-9297/2000/6606-0011$02.00

the use of linkage disequilibrium (LD) and coalescenceanalyses to provide estimated dates for disease-causingmutations. Earlier applications of these methods to AJgenetic diseases demonstrated that 190% of the mutantalleles causing idiopathic torsion dystonia (ITD) arosefrom a single founder mutation (Risch et al. 1995). Therelatively recent coalescence point of the ITD mutation(∼16 generations, corresponding to ∼400 years), coupledwith its relatively high incidence (∼1/300) in the present-day AJ, underscored the role that nonuniform popula-tion expansion has in establishing the current high fre-quency of the ITD disease allele in the AJ. Similaranalyses have been reported for other mutations in theAJ population, including the coagulation factor XI typeII lesion (Peretz et al. 1997; Goldstein et al. 1999), theBRCA1 185delAG deletion (Bar-Sade et al. 1998), andthe APC I1307K mutation (Patael et al. 1999). Prelim-inary studies have suggested that the AJ acid b-glucos-idase (GBA) mutation, N370S, the most frequent causeof Gaucher disease (GD [MIM 230800]) in this popu-

1822 Am. J. Hum. Genet. 66:1821–1832, 2000

lation, shares a common origin with the allele presentin the non-Jewish European population, but these stud-ies have not provided a dating estimate for this mutation(Diaz et al. 1998).

In this study, linkage and haplotype analyses wereused to investigate the origins of the most prevalentGBA mutations in the AJ population: N370S ( )q ∼ .03and 84GG ( ) (Eng et al. 1997). For these stud-q ∼ .003ies, a linkage map of 10 short tandem repeats (STRs)flanking the GBA gene was constructed and then wereused to genotype AJ and non-Jewish N370S and AJ84GG chromosomes. Haplotype and LD analyses in-dicated that the N370S mutation existed in the generalEuropean population prior to the founding, circa the1st millennium, of the AJ population in Europe. Similaranalyses revealed that GD 84GG chromosomes had aunique, conserved haplotype, indicating the occurrenceof an ancestral AJ founder. Coalescence analyses sug-gested that the N370S and 84GG mutations coalescedin the AJ population 48 and 55.5 generations ago(∼1,200–1,400 years before present [YBP]), around thetime that the AJ population became established inEurope.

Subjects, Material, and Methods

Subjects and Sample Collection

Blood was obtained, with informed consent, from pa-tients with GD type 1 and from relevant family memberswith GBA mutations N370S and 84GG who were en-rolled in the study; these individuals included N370Shomozygotes, N370S heterozygotes, and N370S/84GGcompound heterozygotes. By means of the PuregeneDNA isolation kit (Gentra), genomic DNA was isolatedfrom peripheral leukocytes. Phasing of the STR markerson mutation-bearing chromosomes was initially per-formed with a subset of 25 AJ and 29 non-Jewish Eu-ropean N370S chromosomes and 4 AJ 84GG chromo-somes, from families in which parents and/or siblingswere also available. In addition, 60 AJ and 7 Portuguesehomozygotes were studied to determine the allele fre-quencies of the STRs. To obtain allele-frequency dataon the 84GG chromosomes, 32 N370S/84GG com-pound heterozygotes were analyzed. To assess other eth-nic groups for the presence of the common N370S foun-der haplotype, 18 N370S chromosomes from patientsof various ethnic origins were haplotyped, including pa-tients with type 1 GD who were of British ( ), Greekn = 8( ), Macedonian ( ), Sephardic Jewish ( ),n = 4 n = 2 n = 2Algerian ( ), and West Indian ( ) ancestry.n = 1 n = 1

On the basis of preliminary findings, a second set ofAJ subjects, including 58 N370S homozygotes and 36N370S/84GG compound heterozygotes, were genotypedfor coalescence analyses. Analysis of the first set of sub-

jects revealed that the most informative markers for LDanalysis of N370S chromosomes were D1S2624 andD1S1600; therefore, only these markers were typed forthe additional N370S homozygotes. Similarly, initialresults indicated that markers D1S305 and D1S2715were most informative for 84GG chromosomes, so onlythese markers were typed for 84GG/N370S compoundheterozygotes.

Linkage Mapping

Searching the public databases of the Whitehead In-stitute for Biomedical Research/MIT Center for GenomeResearch and The Cooperative Human Linkage Centeridentified maximally informative polymorphic STRmarkers mapping to 1q21; these included D1S305,D1S1595, D1S1600, D1S1653, D1S2128, D1S2624,D1S2715, D1S2721, D1S2777, and the trinucleotide-repeat marker PKLR. For linkage analysis, genotypedata for these markers were obtained from 61 of theCEPH reference families, representing 1,036 meioses.STRs were PCR-amplified under standard conditions(Hudson et al. 1995)—except for markers D1S1595 andD1S1600, which required annealing temperatures of617C and 507C, respectively. The map order and distancebetween the markers were determined by use of theMultiMap software package (Matise et al. 1994).

Analysis

To estimate allele frequencies on population-specificcontrol chromosomes, as required for LD analysis, 83AJ chromosomes (39 from family members of patientswith GD plus 44 from unrelated AJ families), 63 Frenchchromosomes (15 from unaffected family members plus48 from French CEPH families) and 61 Portuguese chro-mosomes (25 from unaffected family members plus 36from random unrelated individuals) were surveyed.Allele frequencies were determined by direct allelecounting.

Estimates of marker-allele frequencies for N370Schromosomes from French and Portuguese subjects alsocould be obtained directly by allele counting, and var-iances of these frequencies were obtained with the for-mula where p is the observed frequency andp(1 2 p)/N,N is the number of alleles. The situation was more com-plicated for the AJ population, because of the inclusionof the N370S/84GG compound heterozygotes. In thiscase, the EM algorithm was used to obtain maximum-likelihood estimates (MLEs) of the allele frequencies(Ceppellini et al. 1995; also see the Appendix). Signifi-cance of LD was then assessed by comparison of allelefrequencies of mutation-bearing chromosomes versusthose of control chromosomes. Specifically, let pI and jp

be the allele frequency and its standard error for allelei on mutation-bearing chromosomes, and let ri and jr be

Diaz et al.: Gaucher Disease N370S and 84GG Mutations 1823

the corresponding frequency and standard error for con-trol chromosomes. Then, statistical significance can bederived by the statistic . Because2 2ÎZ = (p 2 r)/ j 1 ji i p r

multiple alleles were tested at a locus, a statistical cor-relation for multiple tests was employed. A Bonferronicorrection was applied by multiplication of the obtainedsignificance level, P, for a one-sided test (since each allelewas tested for evidence of increased frequency on mu-tation chromosomes) by the number of alleles tested.Every allele with a frequency 15% on the mutation chro-mosomes was tested, and, therefore, nominal P valueswere multiplied by the number of such alleles, to obtaina final P value. This correction may be conservative,especially for loci with a large number of alleles. In thesecases, we also considered the anticonservative correc-tion, , where n is the number of alleles tested,n1 2 (1 2 P)to determine a range for the actual P values

The degree of LD at loci flanking the conserved coremarkers was assessed by calculation of the parameter d

(Bengtsson and Thomson 1981; Risch et al. 1995).Chromosomes bearing the conserved haplotype (definedas markers varying, from the consensus sequence, at nomore than one position) were identified in each popu-lation, and associated allele frequencies were calculatedfor the flanking markers as described above. The d valuewas obtained by use of the formula ,d = (p 2 r)/(1 2 r)where p is the frequency of the associated allele on mu-tation-bearing chromosomes and r is the frequency ofthe same allele on control chromosomes. A confidenceinterval for d was calculated under an assumption ofindependence for the sampled chromosomes. Since d =

is a function of the ratio of two in-1 2 (1 2 p)/(1 2 r)dependent random variables, the variance of d can beapproximated by

21 jp2 2 2( ) ( )Var d = j j 1 1 2 p 1 .( )r p4 2( ) ( )1 2 r 1 2 r

An approximate confidence interval of d can then beobtained as , where .2d 1 1.96j j = Var(d)d d

To estimate the coalescence time for the founder mu-tations, or the number of generations to the most recentcommon ancestor (MRCA), the formula used was

lndG = , (1)

ln(1 2 v)

where v is the recombination fraction with regard to themarker analyzed (Risch et al. 1995). For calculation ofG, the GBA locus was assumed to colocalize withD1S2777, since the degree of LD observed on N370Sand 84GG chromosomes was generally highest at thismarker. A minimum confidence interval for G was cal-culated by use of the formula given above, with theminimum confidence interval for d.

It is difficult to calculate a confidence interval for theestimate of G, because, under the assumption of LD, themutation-bearing chromosomes are not independent(Risch et al. 1995; Rannala and Slatkin 1998). The de-gree of nonindependence—and, hence, the confidenceinterval—is strongly influenced by assumptions regard-ing the demographic history of the population. Rapidexpansion after initial introduction of the mutation leadsto greater statistical independence and smaller confi-dence intervals, whereas constant population size leadsto less statistical independence and broader confidenceintervals (Rannala and Slatkin 1998; Goldstein et al.1999). Here we provide a minimum confidence intervalfor G, using the calculated 95% confidence interval ford and equation (1). We note that this interval is appro-priate only for a rapidly expanding mutation (which isnot unlikely for the mutations under consideration here)and that the true confidence interval may be substantiallylarger. A more refined estimate of G, by examination ofmultiple markers simultaneously, might be possible intheory; however, in this case, because most of the chro-mosomes are unphased, it is not clear that such wouldbe easily implemented or of great benefit.

Results

Genotyping Reveals a Founder N370S Haplotype

To facilitate haplotype and LD analyses of the GDN370S and 84GG mutations, a genetic map of poly-morphic STR markers flanking the GBA gene was con-structed (fig. 1). The genetic map spanned 9.3 cM andcontained 10 STRs, several of which were at no meas-urable recombination distance from each other. TheGBA gene was positioned relative to the mapped STRs,on the basis of available physical-mapping and sequencedata for the GBA region and for the relevant STRs. The5′ end of GBA is 71 kb upstream of the 5′ end of theliver-specific pyruvate kinase gene (Demina et al. 1998).An intragenic trinucleotide repeat in intron 11 of thisgene, PKLR (Lenzner et al. 1994), colocalized withmarkers D1S1595 and D1S2721 on the genetic linkagemap. Physical-mapping data placed GBA on YACs con-taining marker D1S2777. Thus, the GBA locus lies veryclose to the four central markers PKLR, D1S1595,D1S2721, and D1S2777.

For the N370S mutation, 54 homoallelic or hetero-allelic AJ individuals in 19 families were genotypedfor the 10 mapped STRs. Haplotype analysis of thesefamilies with GD revealed a highly conserved 6-5-4-6haplotype for the markers PKLR-D1S1595-D1S2721-D1S2777, respectively (table 1). Additional genotypingof 61 AJ patients who were homoallelic for the N370Smutation demonstrated conservation of the ancestralhaplotype at three or all four of these core STRs, in 133

1824 Am. J. Hum. Genet. 66:1821–1832, 2000

Figure 1 Linkage map of the GBA region. The order of 10 poly-morphic STRs flanking the GBA gene on 1q21 was determined bylinkage mapping using families from the CEPH panel. Markers arearrayed from centromere (top) to telomere (bottom), with the obtainedrecombination distances shown. The position of the GBA gene as de-rived from physical mapping, sequence, and LD data is indicated.

Table 1

Allele-Frequency Estimates for N370S, 84GG, and CEPH ControlChromosomes, for the Central Four Markers

MARKER

AND

ALLELEa

FREQUENCY OF CHROMOSOMES

ControlsCEPH)

N370S 84GG

Ashkenazim French Portuguese Ashkenazim

N = 400 N = 117 N = 17 N = 23 N = 37

PKLR:1 .165 .000 .000 .043 .8652 .092 .009 .000 .000 .1353 .362 .009 .118 .000 .0004 .215 .026 .059 .000 .0005 .080 .026 .059 .000 .0266 .078 .932 .706 .783 .0157 .008 .000 .059 .174 .000

N = 388 N = 142 N = 26 N = 39

D1S1595:2 .131 .007 .000 .000 .0003 .062 .000 .000 .077 .0004 .000 .000 .000 .077 .0005 .330 .958 .824 .731 .0787 .294 .028 .118 .115 .9488 .103 .007 .059 .000 .000

N = 374 N = 137 N = 16 N = 26 N = 38

D1S2721:1 .024 .007 .000 .000 .0002 .064 .007 .000 .077 .0003 .219 .029 .038 .077 .0004 .235 .905 .563 .577 .0005 .045 .007 .063 .038 1.0006 .094 .015 .063 .038 .0007 .035 .007 .000 .000 .0008 .275 .022 .313 .192 .000

N = 420 N = 147 N = 17 N = 27 N = 39

D1S2777:2 .000 .000 .000 .000 .0003 .021 .000 .000 .000 .0004 .093 .007 .059 .148 .0005 .198 .014 .059 .111 1.0006 .543 .973 .882 .741 .0007 .064 .007 .000 .000 .000

a For clarity of presentation, uncommon alleles that are not presentin disease chromosomes have been omitted.

(95%) of the total 140 informative AJ N370S alleles,with only seven chromosomes varying at two or moreSTRs (table 2). Of note, the conserved haplotype wasuncommon, found in only 3 (3.6%) of 83 of the controlAJ chromosomes (data not shown). These could also beN370S carriers, since the population frequency of themutation is ∼6% (Eng et al. 1997).

Genotyping of non-Jewish European patients with GDtype 1 who were homoallelic or heteroallelic for theN370S mutation revealed the same founder haplotypein the majority of individuals (table 1). This group in-cluded 13 Portuguese and 12 French families with GD,from which 14 Portuguese and 15 French N370S hap-lotypes were phased. An additional 14 Portuguese and2 French chromosomes from patients homoallelic forN370S also were genotyped. The ancestral 6-5-4-6 hap-lotype was not detected in 44 French control haplotypes(29 derived from French families in the CEPH GenotypeDatabase and 15 derived from unaffected French indi-viduals in this study) or in 25 Portuguese control hap-lotypes, thereby confirming its rarity in these popula-

tions. Thus, a majority of the N370S alleles in AJ andnon-Jewish European populations shared the common6-5-4-6 haplotype.

Among the French and the Portuguese N370S chro-mosomes, the degree of allelic diversity at the conservedcore markers was strikingly greater than that in the AJchromosomes. Approximately 18% (3 of 17) and 33%(10 of 27) of French and Portuguese N370S chromo-somes, respectively, varied from the ancestral 6-5-4-6haplotype at two or more markers, as compared with5% of the AJ N370S chromosomes (table 2). Indeed,the frequency of the associated founder allele for each

Diaz et al.: Gaucher Disease N370S and 84GG Mutations 1825

Table 2

Conservation of N370S Chromosomes from VariousOrigins

POPULATION

NO. OF N370SCHROMOSOMES

% OF N370SCHROMOSOMES

CONSERVEDTotal Ancestrala

Ashkenazi 140 133 95Portuguese 28 18 64French 17 14 82Other 18 15 83

a At least three of the four core haplotype markersare conserved.

of the four loci was significantly lower ( ) in bothP ! .05the French and the Portuguese N370S chromosomesthan in the AJ N370S chromosomes (table 1). This ob-servation could be explained either by greater decay ofthe conserved haplotype in the non-Jewish populationsor by recurrent N370S mutations in different haplo-types. The former possibility would imply that theN370S founder mutation had existed for a longer timein the non-Jewish populations than in the Ashkenazimand suggests the direction of gene flow for this particularallele.

Genotyping also was performed on GD N370S chro-mosomes of British ( ), Greek ( ), Macedoniann = 8 n = 4( ), Sephardic Jewish ( ), Algerian ( ), andn = 2 n = 2 n = 1West Indian ( ) ancestry. Haplotypes were inferredn = 1for 14 of these 18 chromosomes. The remaining fourchromosomes were from two patients—one Greek andone Macedonian—who were homoallelic for N370S.The Greek patient’s genotype was divergent at PKLR,for both alleles (i.e., 3/5) and at D1S1595 for one allele(i.e., 5/2); thus, one chromosome had the ancestral hap-lotype at three of the four conserved markers whereasthe other had only two of the conserved alleles. TheMacedonian patient was heterozygous at all of the mark-ers, consistent with one completely conserved and onecompletely divergent chromosome. Of these 18 N370Schromosomes from diverse ethnic/demographic groups,15 (83%) had the ancestral 6-5-4-6 haplotype, a fre-quency comparable to the 82% of ancestral chromo-somes among the French N370S alleles (table 2).

LD at Markers Flanking the Conserved N370SHaplotype

Table 3 shows the allele frequencies for the STRsflanking the conserved 6-5-4-6 markers for the AJN370S chromosomes bearing the conserved 6-5-4-6haplotype, for all the 84GG chromosomes, and for thecontrol AJ chromosomes. The allele frequencies formutation-bearing chromosomes were compared withthose for control chromosomes, to assess LD. For

N370S, four markers showed significant LD. For threeof the markers, a single allele was at significantly in-creased frequency (D1S2715, allele 9, , cor-Z = 4.11rected ; D1S2624, allele 5, , correctedP ! .001 Z = 8

; and D1S1600, allele 2, , correctedP ! .001 Z = 7.11). For D1S305, the situation was more compli-P ! .001

cated; three alleles (alleles 4, 5, and 6) were at increasedfrequency, although only the increase in frequency ofalleles 5 and 6 was statistically significant (allele 5,

, corrected ; and allele 6, , cor-Z = 2.94 P ! .01 Z = 3.16rected ); however, it is likely that either the 4 alleleP ! .01or 5 allele is the ancestral allele, as detailed below.

Coalescence Estimates for AJ N370S Chromosomes

The presence of detectable LD at markers a definedmap distance from the GBA locus permitted the esti-mation of the coalescence times for the founder N370Sand 84GG mutations in the AJ population. The d valuesgiven in table 3 (for all markers except D1S305 forN370S) and the recombination fractions from table 4were used to calculate the values of G, the number ofgenerations to the MRCA, for each mutation. As notedabove, several alleles at marker D1S305 appeared to beassociated with the N370S mutation. Therefore, allN370S chromosomes with the disease-associated allele9 at D1S2715 (by phasing or homozygosity) were iden-tified, and the frequencies of the various D1S305 alleleswere calculated for this subset of 69 chromosomes. TheD1S305 allele frequencies were as follows: allele 4,63.8% ( ); allele 5, 23.2% ( ); allele 6, 3.0%n = 44 n = 16( ); and allele 8, 10.1% ( ). Two alleles hadn = 2 n = 7significantly increased frequencies: allele 4 ( ,Z = 2.05

) and allele 5 ( , ). In this subset,P ! .05 Z = 2.59 P ! .01allele 6 did not have increased frequency. These resultssuggested either that allele 4 or allele 5 was the progen-itor or that an early recombination or mutation eventhad led to an increased frequency of both of these alleleson N370S chromosomes. Therefore, a d value was cal-culated, by combining of both of these alleles, leadingto a d estimate of [.516 1 .195 2 (.459 1 .071)]/[1 2

(table 4). As noted in table 3, allele(.459 1 .071)] = .3854 at D1S305 and allele 9 at D1S2715 were very commonin the AJ control chromosomes. Thus, only markersD1S2624 and D1S1600 were typed in the follow-up setof N370S chromosomes.

To account for the uncertainty inherent in the map-ping of tightly linked loci, the v values from the presentstudy were compared with those from the genetic andintegrated maps publicly available from the Center forMedical Genetics, Marshfield Medical Research Foun-dation and The Genetic Location Database of the Uni-versity of Southampton (table 4). The fact that the me-dian v values at three of the four markers used forcoalescence analysis were identical to the v values ob-

1826 Am. J. Hum. Genet. 66:1821–1832, 2000

Table 3

Allele-Frequency Estimates for N370S, 84GG, and Normal Control Chromosomes, with Associated d Values

LOCUS

AND ALLELE

FREQUENCY (d) IN

AJ Population French Population Portuguese Population

Control N370S 84GG Controls N370S Controls N370S

N = 85 N = 193 N = 58 N = 66 N = 16 N = 24 N = 27

D1S305:3 .094 .017 .100 .015 .000 .083 .0004 .459 .516 (.105) .154 .439 .438 (0) .375 .185 (!0)5 .071 .195* (.133) .059 .015 .125 (.112) .042 .000 (!0)6 .024 .121* (.099) .027 .152 .313 (.190) .125 .259 (.153)8 .235 .106 .588 (.461) .227 .125 .167 .0569 .094 .031 .052 .121 .000 .167 .33310 .012 .010 .020 .015 .000 .041 .000

N = 83 N = 182 N = 53 N = 58 N = 16 N = 51 N = 26

D1S2715:2 .000 .006 .452** (.452) .103 .063 .118 .0773 .108 .015 .165 .052 .063 .059 .0384 .301 .133 .130 .293 .313 .353 .3085 .012 .017 .000 .034 .000 .039 .0006 .060 .060 .000 .000 .063 .000 .0008 .024 .006 .019 .069 .063 .078 .0009 .422 .688** (.460) .099 .293 .375 (.120) .333 .500 (.250)10 .024 .076 .098 .317 .125 .020 .000

N = 83 N = 268 N = 32 N = 60 N = 16 N = 50 N = 24

D1S2624:1 .024 .000 .000 .050 .125 .000 .0002 .313 .134 .088 .250 .438 .300 .2083 .036 .008 .000 .067 .000 .100 .0424 .373 .198 .912** (.860) .267 .313 .260 .4585 .205 .631** (.536) .000 .317 .125 (!0) .320 .291 (!0)6 .048 .030 .000 .000 .000 .000 .000

N = 75 N = 258 N = 32 N = 61 N = 17 N = 24 N = 24

D1S1600:2 .067 .326** (.316) .000 .033 .000 (!0) .208 .042 (!0)3 .427 .207 .394 .344 .294 .333 .4584 .333 .346 .368 .410 .412 .167 .3335 .173 .075 .238 (.079) .180 .294 .125 .167

* .P ! .01** .P ! .001

tained in the present study supports the overall accuracyof the current linkage map. The age estimates for N370S,for D1S305, D1S2715, D1S2624, and D1S1600, variedwith the genetic map used but ranged from a minimumof 17 generations (marker D1S2624) to a maximum of136 generations (marker D1S305), both estimates beingbased on the Southampton map. When the medians ofthe v values from the three maps were used, the rangewas 32–68 generations. The minimum confidence inter-vals when median v values were used were 37–123 gen-erations for D1S305, 34–84 generations for D1S2715,31–53 generations for D1S2624, and 26–40 generationsfor D1S1600. The latter two were based on substantially

larger sample sizes. The median value for G was 48generations. If it is assumed that a generation comprises25 years, the coalescence time for this mutation was1,200 YBP, or approximately year 800 of the commonera (C.E.). It is noteworthy that the founding of the AJpopulation at ∼900 C.E., when several thousand Jewsmigrated into the Rhineland and subsequently expandedinto a much larger population (Weinryb 1972), fallswithin the confidence intervals for these estimates of thecoalescence. We note that our analysis did not includemutation at the markers (e.g., see Goldstein et al. 1999).Presumably, the marker mutation rate is substantiallyless than the recombination rate considered here (i.e.,

Diaz et al.: Gaucher Disease N370S and 84GG Mutations 1827

Table 4

Estimates of Coalescence Generations for N370S and 84GG Chromosomes in the AJ Population.

MARKER

N370S 84GG

d

G (v)a

d

Ga

Marshfield SouthamptonPresentStudy Median Marshfield Southampton

PresentStudy Median

D1S305 .385 56 (.017) 136 (.007) 68 (.007) 68 (.014) .461 45 110 55 55D1S2715 .460 45 (.017) 111 (.007) 55 (.007) 55 (.014) .452 46 113 56 56D1S2624 .536 41 (.015) 17 (.036) 69 (.009) 41 (.015) .860 10 4 17 10D1S1600 .316 38 (.030) 29 (.039) 32 (.035) 32 (.035) .079 83 64 71 71

a Recombination distances Marshfield = Center for Medical Genetics, Marshfield Medical Research Foundation; Southampton = TheGenetic Location Database of the University of Southampton. v Values are from the various maps, under the assumption that GBAcolocalizes with D1S2777.

1.5%–3.5%). Inclusion of mutation within the coales-cence analysis would lead to coalescence times slightlymore recent than those given in table 4.

LD Analysis of the Non-Jewish N370S Chromosomes

The degree of LD with associated alleles on N370Sfounder chromosomes in the AJ population can be con-trasted with that in the French and Portuguese popu-lations (table 3). At D1S2715, the associated allele 9,which has a d value of .46 in the AJ population, wasalso associated in the French and Portuguese popula-tions, but to a lesser extent ( for the Portuguesed = .25population, and for the French population), andd = .12the difference was statistically significant for each (P !

). At D1S2624, the associated allele 5 allele was not.05increased in the Portuguese or the French samples (d !

, ). Finally, at D1S1600, the associated allele 20 P ! .05allele was not increased in either the Portuguese or theFrench populations (again, ). Indeed, no otherP ! .05consistent allelic associations at any of these markerswere present in the Portuguese and the French popula-tions. This could not be due to the smaller sample sizeof the French and the Portuguese populations, since acomparable level of LD (i.e., or .536) would haved = .46also been statistically significant in those populations.Thus, there was minimal LD on N370S progenitor chro-mosomes at 0.9%–1.5% recombination distance inthe non-Jewish European populations. These resultsstrongly suggest that the origin of the founder N370Smutation was more widespread and more ancient thanit would be if it had occurred with the founding of theAJ population.

Although the absence of significant LD in the non-Jewish European population precluded direct coales-cence analysis, it was possible to estimate the relativeage of the N370S mutation in this population. The ageof a mutation in a population is proportional to logd,so that, even in the absence of known v values, the ratioof ages can be computed by the ratio of the logd values.

Table 5 show, for the four conserved markers, the d

values and the ratios of logd values in the AJ populationversus those in the French or the Portuguese populations.For the French population, the median R value is 5.0;for the Portuguese population, the median R value is7.0; and the median R value for both populations is 5.6.Assuming an age of 1,200 years for the AJ populationgives an estimate age of ∼6,700 years for the non-JewishEuropean N370S mutation. This value may be a slightoverestimation, because of the inclusion of de novo mu-tations, as discussed below. Nonetheless, it suggests thatlittle LD existed at the flanking loci in the extant pop-ulation at the time of the founding of the Ashkenazimand that survival of a single AJ N370S chromosomefrom the founding has led to LD at the adjacent markers.

Evidence of De Novo N370S Mutations

The allelic diversity at the conserved N370S 6-5-4-6haplotype, especially in the non-Jewish populations,raised the possibility of distinct, recurrent mutations.Thus, N370S chromosomes carrying atypical haplotypeswere screened for the Pv1.1 intragenic polymorphism(Horowitz et al. 1989). This marker is one of 12 intra-genic polymorphisms in the GBA gene that form twohaplotypes (i.e., 1 and 2) that are in complete LD witheach other. The (2) haplotype occurs at a frequency of∼.7 in the general population and has been invariablyassociated with the N370S mutation, a finding that issuggestive of a common founder mutation for the N370Sallele (Zimran et al. 1990; Beutler et al. 1992). Of the23 atypical chromosomes, defined as having two or morevariant alleles at the conserved 6-5-4-6 haplotype, 3 hada Pv1.1(1) allele (table 6), which was confirmed by di-rect sequencing (not shown), providing evidence of re-current N370S mutations. If a proportion, q, of these23 atypical chromosomes were nonancestral, then .3qwould be the frequency of the Pv1.1(1) alleles amongthese chromosomes. Thus, , so that we.3q = 3/23 = .13can estimate q as .43; that is, 43% of the atypical hap-

1828 Am. J. Hum. Genet. 66:1821–1832, 2000

Table 5

d Values and Ratio, R, of logd, for French andPortuguese N370S Chromosomes, Compared withThose of AJ Chromosomes

MARKER

d (R) FOR

AJPopulation

FrenchPopulation

PortuguesePopulation

PKLR .926 .681 (5.0) .765 (3.5)D1S1595 .937 .737 (4.7) .598 (7.9)D1S2721 .876 .429 (6.4) .447 (6.1)D1S2777 .941 .742 (4.9) .433 (13.8)

Table 6

Atypical N370S Chromosomes

POPULATION

ALLELE(S) FOR MARKERPv1.1

ALLELE(S)PKLR D1S1595 D1S2721 D1S2777

Ashkenazim 6 7 7/4 4 24 8 8/3 6 25 5 3 6 2

3/4a 5/7 5/8 6/6 2/24/2a 7/5 8/1 6/6 1/1

Portuguese 6 5 8 4 27 4 4 6 27 5 8, 5 4 2

No data 5/5 8/8 5/5 2/26/6a 7/7 2/2 4/6 2/26/1a 3/5 8/3 5/4 2/2

French 5 5 8 6 26 7 8 6 27 5 8/5 4 2

Macedonian 5/6a 8/5 3/4 5/6 1/2Greek 3/5a 2/5 4/4 6/6 2/2English 6 7 8 7 22

a Unphased genotype data from N370S homozygotes.

lotypes carry de novo mutations, while the remainderrepresent recombination or marker mutation events.

Haplotype and LD Analyses for the 84GG Mutation

The presence of a conserved haplotype at the markersPKLR-D1S1595-D1S2721-D1S2777 was apparent fromexamination of the N370S/84GG genotype data. At eachmarker, two alleles were much more frequent than theywere in control chromosomes, with the first allele cor-responding to the conserved N370S allele and with thesecond allele defining the conserved 84GG allele. Theallele frequencies given in table 1 were calculated bydirect counting of the non-N370S allele at each marker,except in two individuals, one of whom was heterozy-gous for a nonconserved N370S allele and the other ofwhom was heterozygous for a conserved 84GG allele atsingle markers. These alleles defined a 1-7-5-5 haplotypeat the markers PKLR-D1S1595-D1S2721-D1S2777 forthe 84GG chromosomes. Phasing of four 84GG chro-mosomes demonstrated that three of these chromosomeshad the completely conserved haplotype and that thefourth had a single variation (i.e., allele 2 at PKLR).Thus, the haplotypes from the phased chromosomesconcurred with the haplotype predicted by the allele-frequency estimates.

The allele frequencies at nonconserved markers in the84GG chromosomes were estimated by maximum like-lihood (see Subjects, Material, and Methods). As seenin table 3, three markers flanking the conserved hap-lotype showed significant LD, which, in each case, wasattributable to a single allele (D1S305, allele 8, Z =

, corrected ; D1S2715, allele 2, ,4.16 P ! .001 Z = 6.52corrected ; and D1S2624, allele 4, ,P ! .001 Z = 7.17corrected ). At D1S1600, one allele (allele 5)P ! .001was modestly increased but did not reach statisticalsignificance.

Coalescence Analysis of the 84GG Mutation

Coalescence analysis of the 84GG mutation gave sim-ilar age estimates for the two proximal markers, D1S305and D1S2715 (table 4), but gave somewhat disparate

results for the distal markers, D1S2624 and D1S1600.When median v values were used, the minimum confi-dence intervals for D1S305, D1S2715, and D1S1600were overlapping (30–93, 38–81, and 33–` generations,respectively). The interval for D1S2624 (0–25 genera-tions) did not overlap the minimum intervals for theother three markers, perhaps reflecting the fact that theseare minimal intervals and do not account for potentialerrors in v. A median value of G for the 84GG chro-mosomes was ∼55.5 generations (or ∼1,400 YBP), sim-ilar to that observed for N370S. This is probably a rea-sonable estimate, because marker D1S2715 ( )G = 56provided the greatest accuracy, since allele 2 on 84GGchromosomes was absent on control chromosomes andwas nearly absent on N370S chromosomes. In contrast,the associated D1S2624 allele 4 ( ) was very com-G = 10mon (frequency .373) on control chromosomes.

Discussion

In the present article, we have reported detailed hap-lotype and LD analyses that build on our earlier work(Diaz et al. 1998), in an attempt to determine the genetichistory of the GBA N370S and 84GG mutations causingGD. Previous studies of the intragenic GBA polymor-phism, Pv1.1 in intron 6, found that all N370S chro-mosomes examined occurred on the Pv1.1(2) haplotype(Zimran et al. 1990; Beutler et al. 1992; Lacerda et al.1994a). The results of those studies were suggestive ofa common founder but were not conclusive, since the(2) haplotype was found to have a population frequencyof ∼.7 (Beutler et al. 1992; Lacerda et al. 1994a). In

Diaz et al.: Gaucher Disease N370S and 84GG Mutations 1829

contrast, the results described here clearly demonstratethat most N370S chromosomes derive from an ancestralfounder. In addition, our study has documented the ex-istence of de novo N370S mutations occurring on aPv1.1(1) haplotype. Such alleles are uncommon, sinceonly three Pv1.1(1) alleles were identified in the 303N370S chromosomes haplotyped. The mechanism of re-currence is unclear, since the mutation does not occurat a CpG dinucleotide; nor are degenerate repeat motifspresent that could contribute to replication errors.

Because of its high prevalence in the AJ population,N370S has been considered to be a Jewish mutation. Itwas speculated that its presence in non-Jewish Europeanpopulations might have resulted from AJ admixture(Lacerda et al. 1994b). However, the results of LD anal-yses reported here, of N370S chromosomes with theancestral haplotype from AJ and from non-JewishFrench, Portuguese, and other ethnic groups, are notconsistent with gene flow from the AJ into the non-AJpopulation. Despite the fact that the conserved fourhaplotype markers tightly linked to the GBA locus showvery strong disequilibrium for the N370S chromosomes,only the AJ chromosomes show detectable LD at themore distant markers on the 9.3-cM map. The increasedallelic diversity in non-Jewish N370S chromosomescompared with that in the AJ N370S chromosomes isconsistent with a more ancient coalescence point ofN370S in the non-Jewish European populations. Theage estimate (∼6,700 YBP) of the non-Jewish EuropeanN370S mutation, calculated by means of the ratio ofthe logd values, suggests that little LD existed at theflanking markers in the extant population at the timeof the Ashkenazi founding and that survival of a singleAJ N370S chromosome from the founding has led toLD at these adjacent markers.

Although confidence intervals are given for the esti-mates of the mutation coalescence in this study, thereare several additional sources of uncertainty that areworth noting. First, map distances and even somemarker orders varied among the available genetic maps.This is not surprising, since the short distances (andcorrespondingly small number of observed recombinantevents between markers) can be readily altered by small-sample bias or by single genotyping errors. Second, theposition of the GBA gene relative to the four conservedmarkers is not precisely known. The G- estimate cal-culations reported here assume that GBA colocalizeswith D1S2777, for two reasons: (1) the d values for theAJ and the French N370S chromosomes and for the AJ84GG chromosomes were maximal at this marker, and(2) instances of alternative alleles for D1S2777 onN370S chromosomes were generally consistent withmutation at the marker (i.e., the flanking markers wereconserved), whereas this was less true for the other threecentral markers. Third, coalescence analysis does not

distinguish recombination between markers from mu-tations of a marker. High mutation frequencies, whichlead to the allelic variability that make these markerspolymorphic, tend to reduce d values and to result in amore ancient estimate for the coalescence time. Even ifwe account for these sources of uncertainty, it is strikingthat the confidence intervals for the coalescence timesof the N370S and 84GG mutations include the ap-proximate time of the Ashkenazi founding, supportinga founder effect to explain the prevalence of these mu-tations in the population.

Our mutation-dating results are widely discrepantwith those in a recent letter comparing haplotypes ob-served in AJ and Spanish patients with GD and datingthe N370S mutation in the Jewish population to aninterval ∼4,200–9,500 YBP (Diaz et al. 1999). Althoughsome variation can be attributed to the use of differentmap distances, this large discrepancy suggests the ex-istence of a calculation error. When the data presentedin that report and the median recombination distancedescribed above were used to estimate G, a value of

generations was obtained, in close agreementG = 50with the results of the present study.

Several disease mutations present in the AJ can betraced back to Middle Eastern progenitor Jewish pop-ulations. These include the factor XI type II mutation(Peretz et al. 1997; Goldstein et al. 1999), the BRCA1185delAG mutation (Bar-Sade et al. 1998), and the APCI1307K mutation (Patael et al. 1999). These examplesappear to be ancient mutations that are unique to theJewish people. In contrast, the results reported here in-dicate that the N370S founder haplotype exists widelyin non-Jewish European chromosomes. The distributionof the N370S mutation is more like that of the factorII polymorphism, G20210A, which is associated withan increased risk of thromboembolism. The G20210Aallele is found at high prevalence among the AJ but isalso found at lower levels in various non-Jewish pop-ulations (Zivelin et al. 1998). The presence of this factorII mutation in multiple European populations, in con-junction with its absence in non-European populations,suggests that it originated in an ancient progenitor pop-ulation postdating the divergence of European andAsian populations. The same inference can be made forthe N370S mutation, which has not been found in anyAsian populations studied to date (Kim et al. 1996;Choy et al. 1997; Eto and Ida 1999).

The underlying genetic mechanism(s) accounting forthe high prevalence of the N370S and 84GG mutationsin the AJ population, as well as for other common AJmutations, has been the subject of considerable debate.The relatively high prevalence of several lysosomal dis-orders in the AJ population, each resulting from severalindependent mutations, suggests the hypothesis thatheterozygote advantage was the operative genetic mech-

1830 Am. J. Hum. Genet. 66:1821–1832, 2000

anism. Although this suggestion is intuitively appealing,a clear biological benefit of heterozygosity for theselysosomal mutations has never been documented. Thefinding that a common N370S allele is concurrentlypresent in the AJ population and in the neighboringnon-Jewish European populations allows some infer-ences to be drawn. To explain the current N370S alleleprevalence among the AJ by heterozygote advantagewould require that selection operate only in this pop-ulation and not on proximate non-AJ populations. Ininstances in which heterozygosity has conferred de-monstrable selective advantage to human populations,by affecting susceptibility to common infectious dis-eases, such uneven selection has not been observed. Se-lection for b-globin mutations, which offer protectionagainst malarial infection (Williams et al. 1996), hasbeen observed in multiple populations inhabiting ma-laria-endemic regions. Similarly, the prevalence of theHIV-coreceptor mutation, CCR5D32, is approximatelythe same in the AJ population as in other, northern-European populations. This mutation, which has nodisease phenotype, is proposed to have recently risen toits current frequency of .10–.13 in northern-Europeanpopulations by conferring immunity to bacterial path-ogens that enter cells by means of the CCR5 receptor(Libert et al. 1998; Stephens et al. 1998). If heterozy-gosity for N370S confers resistance to a common path-ogen, it is unlikely that increased allele prevalencewould be limited to the AJ population.

In contrast to N370S, the 84GG mutation is uniqueto the AJ population and appears to coalesce at ap-proximately the same historical period as does theN370S lesion. If the ancestral alleles for both of thesemutations were extant at the founding of the AJ pop-ulation, ∼1,200 YBP, as suggested by the coalescenceanalyses, then the striking differences in mutation prev-alence in the current AJ population highlight the im-portance of demographic (or epigenetic) factors in theshaping of allele frequencies, at least within this fre-quency range. Although 84GG is a null allele that islethal in homozygosity (Tayebi et al. 1997; R. J. Desnickand C. M. Eng, unpublished results), the negative-selection coefficient is minimal, because of the low fre-quency of other genetic-lethal GD alleles in the AJ pop-ulation. Previous work by Risch et al. (1995) hassuggested that the present-day AJ population is not theproduct of uniform population growth but, rather, isderived primarily from expansion of a subpopulation.The recent finding that the prevalence of Tay-Sachs car-riers among carriers of GD is lower than would be ex-pected on the basis of population-frequency data sug-gests that these mutations have not yet reached geneticequilibrium, a finding that is consistent with the pres-ence of population subdivisions within the AJ popula-tion (Peleg et al. 1998). Thus, a likely explanation for

the currently observed GD allele frequencies appears tobe nonuniform expansion of the AJ population after thefounding bottleneck. In support of this conclusion, com-puter simulations modeling the introduction of diseasealleles into a rapidly growing population have producedresults consistent with skewing of disease-allele fre-quencies, as is seen in the AJ population, in the absenceof selection and in the time frame under discussion (N.Risch, personal communication). Taken as a whole, thefindings in this and other studies indicate the importanceof genetic drift in shaping the disease-allele frequenciesobserved in the contemporary AJ population.

Acknowledgments

The authors would like to thank Drs. Christine Eng, MarieGrace, and Lea Peleg, for additional clinical materials used inthese studies, and Dr. Rina Zaizov, for longstanding activeinvolvement in the clinical and research aspects of this project.These studies were supported in part by a Lucy Moses Foun-dation Fellowship (to G.A.D.) and by NIH grant 5 P30 HD28822 (to the Mount Sinai Child Health Research Center).

Appendix

Estimation of Allele Frequencies in Disease Chromo-somes, by EM Algorithm

Suppose that there are m marker alleles denoted 1–m.Let pi represent the frequency of allele i on N370S chro-mosomes, and let qi represent the frequency of allele ion 84GG chromosomes. Let ai represent the number ofi alleles observed on N370S chromosomes (either fromhomozygotes or from phased heterozygotes), and let bi

represent the corresponding number on 84GG chro-mosomes, and let cij represent the number of ij genotypesfound on the N370S/84GG compound heterozygotes.Let , , and (note that ). TheA = Sa B = Sb C = Sc c = ci i ij ij ji

following recursion formulas can then be used to obtainthe MLEs of pi and qi:

a 1 c O c ∗ p q /(p q 1 p q )i ii ij i j j i i jj(i′p =i A 1 1/2C

and

b 1 c O c ∗ p q /(p q 1 p q )i ii ij i j j i i jj(i′q = .i B 1 1/2C

Standard errors for the estimates of pi and qi can beobtained by standard likelihood theory, by inverting

Diaz et al.: Gaucher Disease N370S and 84GG Mutations 1831

the expected information matrix; for example, for fre-quencies pi and qi, the respective variances, jp andjq, are given by and2 2 2j = q / (q q 2 q ) j =p 22 11 22 12 q

, where2q / (q q 2 q )11 11 22 12

A C(p 1 q 2 2p q )i i i iq = 111 p (1 2 p ) p (1 2 p )i i i i

2C(1 2 2q )i1 ,(p 1 q 2 2p q )i i i i

B C(p 1 q 2 2p q )i i i iq = 122 q (1 2 q ) q (1 2 q )i i i i

2C(1 2 2p )i1 ,(p 1 q 2 2p q )i i i i

and

Cq = .12 (p 1 q 2 2p q )i i i i

Electronic-Database Information

Accession numbers and URLs for data in this article are asfollows:

CEPH Genotype Database, http://www.cephb.fr/cephdbCooperative Human Linkage Center, The, http://lpg.nci.nih

.gov/CHLCGenetic Location Database, The, http://cedar.genetics.soton

.ac.uk/public_html/ldb.htmlCenter for Medical Genetics, Marshfield Medical Research

Foundation, http://www.marshmed.org/geneticsOnline Mendelian Inheritance in Man (OMIM), http://www

.ncbi.nim.nih.gov/omim (for GD [MIM 230800])Whitehead Institute for Biomedical Research/MIT Center for

Genome Research, http://www-genome.wi.mit.edu

References

Bar-Sade RB, Kruglikova A, Modan B, Gak E, Hirsh-YechezkelG, Theodor L, Novikov I, et al (1998) The 185delAGBRCA1 mutation originated before the dispersion of Jewsin the Diaspora and is not limited to Ashkenazim. Hum MolGenet 7:801–805

Bengtsson BO, Thomson G (1981) Measuring the strength ofassociations between HLA antigens and diseases. Tissue An-tigens 18:356–363

Beutler E, West C, Gelbart T (1992) Polymorphisms in thehuman glucocerebrosidase gene. Genomics 12:795–800

Bonne-Tamir B, Adam A (1992) Genetic diversity among Jews.Oxford University Press, New York

Ceppellini R, Siniscalco M, Smith CAB (1955) The estimation

of gene frequencies in a random-mating population. AnnHum Genet 20:97–115

Chase GA, McCusick VA (1972) Controversy in human ge-netics: founder effect in Tay-Sachs disease. Am J Hum Genet24:339–340

Choy FY, Humphries ML, Shi H (1997) Identification of twonovel and four uncommon missense mutations among Chi-nese Gaucher disease patients. Am J Med Genet 71:172–178

Demina A, Boas E, Beutler E (1998) Structure and linkagerelationships of the region containing the human L-type py-ruvate kinase (PKLR) and glucocerebrosidase (GBA) genes.Hematopathol Mol Hematol 11:63–71

Diaz A, Montfort M, Cormand B, Zeng B, Pastores GM, Cha-bas A, Vilageliu L, et al (1999) Gaucher disease: the N370Smutation in Ashkenazi Jewish and Spanish patients has acommon origin and arose several thousand years ago. AmJ Hum Genet 64:1233–1238

Diaz GA, Risch N, Nygaard T, Maire I, Poenaru L, CaillaudC, Sa Miranda C, et al (1998) Gaucher disease: the Ash-kenazi Jewish N370S mutation occurred on an ancient Eu-ropean haplotype. Am J Hum Genet Suppl 63:A211

Eng CM, Schechter C, Robinowitz J, Fulop G, Burgert T, LevyB, Zinberg R, et al (1997) Prenatal genetic carrier testingusing triple disease screening. JAMA 278:1268–1272

Eto Y, Ida H (1999) Clinical and molecular characteristics ofJapanese Gaucher disease. Neurochem Res 24:207–211

Goldstein DB, Reich DE, Bradman N, Usher S, Seligsohn U,Peretz H (1999) Age estimates of two common mutationscausing factor XI deficiency: recent genetic drift is not nec-essary for elevated disease incidence among Ashkenazi Jews.Am J Hum Genet 64:1071–1075

Goodman RM (1979) Genetic disorders among the Jewishpeople. Johns Hopkins University Press, Baltimore

Horowitz M, Wilder S, Horowitz Z, Reiner O, Gelbart T,Beutler E (1989) The human glucocerebrosidase gene andpseudogene: structure and evolution. Genomics 4:87–96

Hudson TJ, Stein LD, Gerety SS, Ma J, Castle AB, Silva J,Slonim DK, et al (1995) An STS-based map of the humangenome . Science 270:1945–1954

Kim JW, Liou BB, Lai MY, Ponce E, Grabowski GA (1996)Gaucher disease: identification of three new mutations inthe Korean and Chinese (Taiwanese) populations. Hum Mu-tat 7:214–218

Lacerda L, Amaral O, Pinto R, Aerts J, Sa MMC (1994a) TheN370S mutation in the glucocerebrosidase gene of Portu-guese type 1 Gaucher patients: linkage to the PvuII poly-morphism. J Inherit Metab Dis 17:85–88

Lacerda L, Amaral O, Pinto R, Oliveira P, Aerts J, Sa MMC(1994b) Gaucher disease: N370S glucocerebrosidase genefrequency in the Portuguese population. Clin Genet 45:298–300

Lenzner C, Jacobasch G, Reis A, Thiele B, Nurnberg P (1994)Trinucleotide repeat polymorphism at the PKLR locus. HumMol Genet 3:523

Libert F, Cochaux P, Beckman G, Samson M, Aksenova M,Cao A, Czeizel A, et al (1998) The dccr5 mutation conferringprotection against HIV-1 in Caucasian populations has asingle and recent origin in Northeastern Europe. Hum MolGenet 7:399–406

Matise TC, Perlin M, Chakravarti A (1994) Automated con-

1832 Am. J. Hum. Genet. 66:1821–1832, 2000

struction of genetic linkage maps using an expert system(MultiMap): a human genome linkage map. Nat Genet 6:384–390

Motulsky AG (1995) Jewish diseases and origins. Nat Genet9:99–101

Myrianthopoulos NC, Naylor AF, Aronson SM (1972) Foun-der effect in Tay-Sachs disease unlikely. Am J Hum Genet24:341–342

Patael Y, Figer A, Gershoni-Baruch R, Papa MZ, Risel S,Shtoyerman-Chen R, Karasik A, et al (1999) Common or-igin of the I1307K APC polymorphism in Ashkenazi andnon-Ashkenazi Jews. Eur J Hum Genet 7:555–559

Peleg L, Frisch A, Goldman B, Karpaty M, Narinsky R, Bron-stein S, Frydman M (1998) Lower frequency of Gaucherdisease carriers among Tay-Sachs disease carriers. Eur J HumGenet 6:185–186

Peretz H, Mulai A, Usher S, Zivelin A, Segal A, Weisman Z,Mittelman M, et al (1997) The two common mutations caus-ing factor XI deficiency in Jews stem from distinct founders:one of ancient Middle Eastern origin and another of morerecent European origin. Blood 90:2654–2659

Rannala B, Slatkin M (1998) Likelihood analysis of disequi-librium mapping, and related problems. Am J Hum Genet62:459–473

Risch N, de Leon D, Ozelius L, Kramer P, Almasy L, SingerB, Fahn S, et al (1995) Genetic analysis of idiopathic torsiondystonia in Ashkenazi Jews and their recent descent from asmall founder population. Nat Genet 9:152–159

Stephens JC, Reich DE, Goldstein DB, Shin HD, Smith MW,Carrington M, Winkler C, et al (1998) Dating the origin ofthe CCR5-D32 AIDS-resistance allele by the coalescence ofhaplotypes. Am J Hum Genet 62:1507–1515

Tayebi N, Cushner SR, Kleijer W, Lau EK, Damschroder-Wil-liams PJ, Stubblefield BK, Den Hollander J, et al (1997)Prenatal lethality of a homozygous null mutation in the hu-man glucocerebrosidase gene. Am J Med Genet 73:41–47

Weinryb BD (1972) The Jews of Poland: a social economichistory of the Jewish community in Poland from 1100–1800.Jewish Publication Society of America, Philadelphia

Williams TN, Maitland K, Bennett S, Ganczakowski M, PetoTE, Newbold CI, Bowden DK, et al (1996) High incidenceof malaria in alpha-thalassaemic children. Nature 383:522–525

Zimran A, Gelbart T, Beutler E (1990) Linkage of the PvuIIpolymorphism with the common Jewish mutation forGaucher disease. Am J Hum Genet 46:902–905

Zivelin A, Rosenberg N, Faier S, Kornbrot N, Peretz H, Mann-halter C, Horellou MH, et al (1998) A single genetic originfor the common prothrombotic G20210A polymorphism inthe prothrombin gene. Blood 92:1119–1124

Zlotogora J, Zeigler M, Bach G (1988) Selection in favor oflysosomal storage disorders? Am J Hum Genet 42:271–273

Zoossmann-Diskin A (1995) ITD in Ashkenazi Jews—geneticdrift or selection? Nat Genet 11:13–15